city research online · this is the published version of the paper. this version of the publication...
TRANSCRIPT
City, University of London Institutional Repository
Citation: Rooney, C., Beecham, R., Dykes, J. ORCID: 0000-0002-8096-5763 and Wong, W. (2017). Dynamic Design Documents for supporting applied visualization. Poster presented at the IEEE VIS 2017, 01 - 06 Oct 2017, Phoenix, USA.
This is the published version of the paper.
This version of the publication may differ from the final published version.
Permanent repository link: http://openaccess.city.ac.uk/19479/
Link to published version:
Copyright and reuse: City Research Online aims to make research outputs of City, University of London available to a wider audience. Copyright and Moral Rights remain with the author(s) and/or copyright holders. URLs from City Research Online may be freely distributed and linked to.
City Research Online: http://openaccess.city.ac.uk/ [email protected]
City Research Online
Dynamic Design Documents for supporting applied visualizationChris Rooney ∗
Middlesex University, UKRoger Beecham †
City, University of London, UKJason Dykes‡
City, University of London, UKWilliam Wong§
Middlesex University, UK
ABSTRACT
A common characteristic of applied visualization is collaborationbetween visualization researcher and domain expert – where the vi-sualization researcher attempts to assimilate sufficient detail arounddata, task and requirements to design a visualization tool that ismanifestly useful. We report on a method for enabling such a col-laboration that can be used throughout the design process to gatherand develop requirements and continually evaluate and support iter-ative design. We do so using highly interactive web-pages that weterm dynamic design documents. Applied during a four-year visualdata analysis project for crime research, these documents enabled aseries of data mappings to be explored by our collaborators (crimeanalysts) remotely – in a flexible and continuous way. We arguethat they engendered a level of engagement that is qualitatively dis-tinct from more traditional methods of feedback elicitation, offereda solution to limited and intermittent contact between analyst andvisualization researcher and speculate that they provided a meansof partially addressing certain intractable deficiencies, such as so-cial desirability-bias, that are common to evaluation in applied datavisualization.
1 INTRODUCTION
Applied or problem-driven visualization usually requires some col-laboration between Information Visualization researcher and front-line analyst [5]. Close engagement with front-line-analysts duringthe early stages of a project allows researchers to learn important de-tail about previously unfamiliar datasets and analysis routines. As anapplied visualization study progresses, front-line-analysts are calledupon to perform analysis, test techniques and evaluate proposedvisual analysis tools against previously identified analysis routinesas designs and requirements develop, data are explored and ideasare generated.
In such examples of ‘visual design methodology’, domain expertstend to contribute heavily during problem specification and visu-alization tool evaluation but less so during visual design. Recentexamples of applied visualization have demonstrated that the dis-tinction between visualization researcher contributing new designsand domain expert contributing subject-matter expertise can be dis-solved [7], with front-line-analysts playing a far more active role inthe design process and the visualization researcher also generatingnew domain-relevant knowledge [3]. Fuller interaction betweendomain expert and visualization researcher, particularly during thedesign phase, has been achieved through, for example, creativityworkshops [2].
Inspired by interactive stories published within data journalism12,we report on an alternative approach whereby design ideas arecommunicated to domain experts using dynamic design documents
∗e-mail: [email protected]†e-mail:[email protected]‡e-mail:[email protected]§e-mail:[email protected]
1http://nyti.ms/2nBpk3o, last accessed 31st March 20172http://bbc.in/2eTznsH, last accessed 31st March 2017
(DDDs). Rather than constraining the process of domain user en-gagement to single events, these highly interactive web-pages canbe navigated by analysts in a setting with which they are familiarand through a process of guided exploration, interaction and play.As well as describing our use and design of DDDs, we reflect onthe level and quality of feedback that was elicited from engaginganalysts in this way.
Combining SPC with Geographic Representations
In the last chapter, we used colour to represent population. In this chapter we usecolour to show signals that occur at the most recent data point (which we will refer toas 'today'). In the first visualisation below, we use a simple mapping to start with - redis any signal over the mean and blue is any signal under the mean. The data span fromJanuary 2011 to December 2016 and 'today' is currently set to 1st May 2015. You canmove the date backward and forward one month at a time using these grey buttons.This allows you to see how the patterns look at different times, with different signalpatterns, and how they develop.
A gapped chart showing whether a signal has beentriggered in a neighbourhood. Click on aneighbourhood to see its SPC chart.The date is 1st May 2015.
(i)
(a)
On a side note, this type of colouring might be useful in the background of some othervisualisation (such as icons representing a signal), so we've implemented a slider bar tocontrol the opacity. How faint can you make the visualisation such that you can stilldetermine the NPUs?
Small multiples with gaps representing the WestMidland neighbourhoods, coloured by NPU.Opacity: 0.67
Willenhall South, Walsall Population: 15818
(ii)
(iii)
(b)
Further to this, we can add the mark indicating the mean of the data point or signal.We adjust the opacity to allow the glyphs to be seen on top of the processes and signals.
A gapped chart showing both signal and processinformation for each neighbourhood, as well as a markshowing the mean of either the last data point or thesignal. Click on a neighbourhood to see its SPC chart.The date is 1st May 2015.
(c)
Figure 1: Three excerpts from our interactive documents. Interactivefeatures include (i) buttons inline with text for controlling dateranges, (ii) hover for additional information, and (iii) a slider tocontrol opacity levels.
2 DYNAMIC DESIGN DOCUMENTS
An excerpt from our DDDs is in Figure 1 and a full set of documentscan be viewed at https://rooch84.github.io/spc/.
Our DDDs communicate an ensemble of redesigns, or visual-ization ‘make-overs’, of Statistical Process Control Charts (SPC).SPC monitoring is an established technique that combines statisticaltheory and visual methods to identify exceptional activity, in thiscase exceptional levels of crime given historical data. The proposedredesigns allow our collaborators (crime analysts) to identify geo-spatial patterns in the ‘signals’ implied by SPCs. They do so byabstracting some of the detail from default SPC charts with whichanalysts are familiar (Figure 1) and by carefully mapping key dataproperties – SPC signals and processes – to visual channels.
In traditional applied visualization, researchers tend to presentincremental versions of self-contained tools. When demonstratingthese tools to domain collaborators, it is very easy to focus on inter-action features and usability concerns instead of design. Our DDDsaim to articulate the design process as a narrative, demonstratingdeficiencies in existing data mappings as a means of explaining andjustifying often new and unfamiliar encodings. We append both newand modified designs to the narrative, providing a provenance trailfor our redesign.
A challenge in applied visualization is around persuading front-line analysts of a new visual grammar, where the benefits conferredby that new encoding may not be obvious [5]. By exploring theredesigns independently using DDDs, we hoped that analysts wouldunderstand and form a view of the new data mappings, as well asdevelop a sensitivity to the trade-offs involved in engineering greaterdata density.
Since spatial analysis of SPC signals motivated the redesign,our DDDs first demonstrated and encouraged front-line analyststo explore the spatial layouts used to overlay neighbourhood-levelsignals in their approximate spatial location. Our DDDs supportedde facto interactions, such as hover for additional information. Bymousing over individual spatial units and visually scanning acrossconventional and abstracted spatial layouts, front-line analysts wereable to learn and also critique our new layout, which was derivedalgorithmically [4].
A further ambition was to encourage analysts to reason criticallyabout SPC methodology given our redesigns. We engineered veryparticular interactions in order to effect this thinking – placing textualoverlays and interactions at certain key moments of our designdescriptions, styling these sections of the text in a way that suggestedinteractivity [1].
We used the d3 visualisation library for implementing our designs.Our intention was to create interactive markdown using Glasseye3,but a lack of control over the visualizations led us to create our owninteractive documents manually using HTML, CSS, and JavaScript.We use Edward Tufte-inspired CSS styling4 to our documents [6],which can be seen in Figure 1. This styling provides clarity andcarries authority, whilst suggesting some informality that we hopedwould encourage experimentation and feedback.
3 DISCUSSION
We found engaging analysts through the use of DDDs to be highlyinstructive. Our collaborators commented that the descriptions of de-sign rationale accompanied with interactive examples ‘demonstratedthe thinking process’. That these documents made transparent theincremental nature of our re-designs, and the genesis of specificdesign decisions, was a particularly positive outcome. Revealingthis process gradually meant that analysts were more confident in in-terpreting more detailed composite views, which they acknowledgedat first glance ‘tend to overwhelm’.
3http://bit.ly/2nRm0hE, last accessed 31st March 20174http://bit.ly/2nJrjmI, last accessed 31st March, 2017
We speculate that the documents enabled a level of engagementthat was qualitatively unique from earlier design study projects inwhich we have participated. By physically separating and distancingresearcher and collaborator, analysts were able to explore designsover several days. This freed analysts from the pressure of provid-ing immediate responses or from exposing the fact that they mightnot immediately understand our redesigns. We detected a slightchange in role – something close to a blurring of ‘domain expert’and ‘visualisation expert’ similar, but in the opposite direction, tothat observed by Wood et al. [7] and in line with the mutual influencedescribed by McCurdy et al. [3]. Perhaps due to a deeper understand-ing of the design process, analysts began to make considered andwell-justified design suggestions rather than simple feature requests.These interventions generated interesting discussion and that suchdiscussion took place is evidence of the level of agency collaboratorsfelt over the designs. Analysts also began to question establishedstandards, using a vocabulary and justification informed by visualdesign principles that were only ever implied by our redesigns.
There was a material difference in the feedback elicited throughanalysts annotating DDDs and from teleconference calls scheduledto discuss our redesigns. In the annotations, analysts expressed well-qualified scepticism around certain abstractions, particularly in ourmore detailed composite views. Such scepticism was not expressedin the teleconference. We speculate that this difference in opinionmay to an extent relate to effects familiar to empirical social science:for example, social-desirability bias, where individuals respond inways that they perceive will be viewed positively by others.
We are still exploring the use of DDDs, but initial experiencessuggest they may help engender focus and substance in appliedvisualization. In particular, we consider them to support tight andeffective working between analyst and visualization researcher.
ACKNOWLEDGMENTS
This research was in part supported by the EU under the EC GrantAgreement No. FP7-IP-608142 to Project VALCRI, awarded to Mid-dlesex University and partners. The authors wish to thank colleaguesat West Midlands Police (UK) for their ideas, continued interest andengagement.
REFERENCES
[1] J. Boy, L. Eveillard, F. Detienne, and J. D. Fekete. Suggested interactiv-ity: Seeking perceived affordances for information visualization. IEEETransactions on Visualization and Computer Graphics, 22(1):639–648,Jan 2016. doi: 10.1109/TVCG.2015.2467201
[2] S. Goodwin, J. Dykes, S. Jones, I. Dillingham, G. Dove, A. Duffy,A. Kachkaev, A. Slingsby, and J. Wood. Creative user-centered visu-alization design for energy analysts and modelers. IEEE Transactionson Visualization and Computer Graphics, 19(12):2516–2525, Dec 2013.doi: 10.1109/TVCG.2013.145
[3] N. McCurdy, J. Dykes, and M. Meyer. Action design research andvisualization design. In BELIV ’16 Proceedings of the Sixth Workshop onBeyond Time and Errors on Novel Evaluation Methods for Visualization,pp. 10–18. Baltimore, ML, USA, 2016.
[4] W. Meulemans, J. Dykes, A. Slingsby, C. Turkay, and J. Wood. Smallmultiples with gaps. IEEE Transactions on Visualization & ComputerGraphics, 23(1):381–390, 2017.
[5] M. Sedlmair, M. Meyer, and T. Munzner. Design study methodology:Reflections from the trenches and the stacks. IEEE Transactions onVisualization and Computer Graphics, 18:2431–2440, 2012.
[6] E. R. Tufte. Visual Explanations: Images and Quantities, Evidence andNarrative. Graphics Press, Cheshire, CT, USA, 1997.
[7] J. Wood, R. Beecham, and J. Dykes. Moving beyond sequential design:Reflections on a rich multi-channel approach to data visualization. IEEETransactions on Visualization and Computer Graphics, 20(12):2171–2180, Dec 2014. doi: 10.1109/TVCG.2014.2346323